OpenAI quietly opens ChatGPT to the public, and the internet rearranges itself overnight

OpenAI did not throw a launch party for ChatGPT. The San Francisco lab, then best known to the technical press for its image generator DALL-E 2, posted a blog entry on the afternoon of 30 November 2022 announcing what it called a "research preview" of a conversational model fine-tuned from its existing GPT-3.5 series. There was no waiting list, no enterprise contract and no press embargo. Anyone with an email address could try it.

The product was built, in OpenAI's own framing, to let researchers and the public probe the limits of large language models. What it actually did was give millions of people their first unsupervised hour with a system that could draft an email, summarise a meeting transcript, write a wedding speech and sketch a Python script in roughly the time it takes a kettle to boil.

From research demo to mass habit

Within five days, OpenAI's chief executive Sam Altman wrote on Twitter that the service had passed a million users. By comparison, Instagram took two and a half months to reach the same threshold and Facebook took ten. As reported by Reuters and later analysed in detail by The New York Times, ChatGPT became the fastest consumer-software adoption curve on record, a distinction it held until Threads briefly took it the following summer.

The system itself was not a meaningful capability jump on the GPT-3.5 base model that had been quietly available to API customers for over a year. The change was packaging. A free chat interface, with a working memory inside the session and a friendly tone, did what a developer playground had not: it made the underlying model feel like a tool a non-technical person could use without instruction.

It was not a new model. It was a new way of meeting one.

An immediate education problem

Schools and universities were the first institutions to feel the consequences. Within ten days, articles in The Atlantic and the BBC were citing teachers who had received essay drafts that were palpably not student work. New York City public schools moved to block the service from their networks in early January 2023. The episode forced a fast public conversation, still unresolved at this writing, about what assessment looks like when a model can produce a passable five-paragraph essay on demand.

On the engineering side, the most striking thing was how quickly the rest of the industry conceded the framing. Microsoft, which had invested in OpenAI a year earlier, accelerated work on what would become Bing Chat. Google, which had developed comparable models internally for years, declared a corporate "code red" within weeks. The product had not changed the technology underneath. It had changed who was allowed to imagine using it.